Andrea Cau 2019 is critical of the LRT

Google translated from Theropoda.blogspot.com March 2019
This just came to my attention, two years after publication by Andrea Cau, PhD. Unfortunately I can’t give credit to the reader who sent me an email with this link because he or she operates under a pseudonym.

“The “The Pterosaur Heresies” site constitutes a sort of “permanent spam” for anyone who wants to search for information and images online relating to fossil amniotes. The author of the site, David Peters, has for years been proposing his own personal and idiosyncratic systematic of tetrapods, which he divulges online through his site, and which is illustrated with a large cladogram having over a thousand taxa included, but which, for today, it has never been published in the technical literature nor is it replicable in any way by third parties, since the author has never published a list of characters or a matrix.”

This is a false statement. Since this online experiment is an ongoing project, taxa and characters are often added on a daily to weekly basis. Scores are changed as experience with new taxa grows. Even so the MacClade matrix (which I have been working with for the past ten years) has always been available by request by clicking on the yellow .nex icon on the webpage that includes the large reptile tree (LRT). Many interested parties have done so over the years.

“Therefore, the phylogeny proposed by Peters is not testable and therefore it is not science, but pseudoscience.”

No. Anyone can test it at any time.

Added a day later: I requested Cau’s .nex file in order to test it. That request was denied this morning. Therefore, ironically, the phylogeny proposed by Cau “is not testable” and quoting Cau himself, “therefore it is not science.” I wouldn’t call it pseudoscience. That would be inappropriate.

“Instead of playing honestly like all of us paleontologists who submit our hypotheses to peer-reviewed scientific journals, Peters is all about rhetoric to self-confirm his own personal scenario of reptilian evolution.”

Not exactly. I have submitted manuscripts to several peer-reviewed academic journals. Nature, Science, Journal of Vertebrate Paleontology, Ichnos, Historical Biology and several other journals (list online here with abstracts) have published my manuscripts over the years.

Added a day later: I report results recovered from software that phylogenetically analyzes input taxa, characters and scores. There is no personal scenario here. I report results, hoping that someday others will duplicate taxon lists. Many have done so. Click here for that list.

“One of Peters’ rhetorical claims is that taxonomic sampling is sufficient to produce a “valid phylogeny”, that is, backing his rhetoric on all other authors, that his phylogeny is the best because it includes a thousand species, against those of other researchers who rarely exceed the hundreds of species analyzed.”

Not exactly. Due to the wide gamut of the taxa included in the LRT, I am often able to indicate a problem in smaller studies that pertinent taxa have been excluded, or irrelevant taxa have been included. Cau’s editorial is either not well researched, because it omits important facts, or it is colored by a rather common irritation that an amateur is pointing out errors of omission in the work of professionals and PhDs.

“The reason why Peters’ phylogeny contains thousands of species is that the list of traits he claims to analyze is very short, with a low number of items analyzed, less than 300. This makes the result of his analyzes extremely doubtful, given that it is extremely unlikely that with not even 300 characters spread over 1000 species, we can have a tree that is completely resolved and devoid of unstable areas.”

Not exactly. The character list is only 235. The fact that each character is multi-state (often up to 10 states) multiplies the efficacy by several magnitudes. The fact is the LRT continues to be able to lump and separate taxa. That’s a fact. The only weakness is my lack of experience with the data and taxa that are added and added and added. Heck, it even works with fish when their bones are provided with tetrapod homologs. Importantly I tend to avoid taxa represented by just a few bones.

“I say this from experience, given that in 15 years I have assembled a matrix of theropods that analyzes more than 1800 characters in over 500 species, and I have found (and like me, all those who carry out serious phylogenetic analyzes with fossils have observed) that all ‘to increase taxa, it is necessary to introduce new characters that describe potentially significant characteristics and which, before the addition of new species, could not be variable (and therefore not significant for identifying clades).”

If Cau is analyzing 500 theropods it is a good guess that some taxa are represented by just a few bones. When a skull taxon nests with a pelvis taxon, then there will inevitably be loss of resolution.

“Obviously, having a few hundred characters, it is quite easy and quick to add new species, since the encodings to be made are relatively few. This is how he gets such a matrix!”

True. When I need more characters, I add more, but I haven’t really needed that to any great extent. Cau is working from a false paradigm. He should test it!

“The situation changes when you have more than 1000 characters, and therefore it takes a lot of work to encode each new species included.”

In my experience, that represents overkill. The degree of precision increases with every new character, but it pretty much levels off with 150 multi-state characters. Thereafter the increase in precision is incremental to miniscule.

“Therefore, since a serious phylogenetic analysis must have a sampling of characters adequate to the number of species, whoever proposes a phylogeny based on not even 300 characters out of over 1000 species, is, in the best of options, lying to himself.”

Testing says otherwise. How can I lie to myself when the software determines relationships that greatly resemble one another? Dr. Cau has turned his ire from taxa and characters to an ad hominem attack. That’s not good science. The scientific method requires that you repeat the prescribed experiment, follow the prescribed methods and report results. Don’t run your own experiment and report that mine doesn’t work based on your results. Suggestions for Dr. Cau: 1) use only every third character to test results. 2) delete incomplete taxa to test results. 3) Examine the pertinent specimens first hand. 4) Provide tracings and reconstructions so readers know the basis for your scores.

“Peters declares that his hypothesis is testable by anyone: listening to him, it is enough to assemble a phylogeny of reptiles, without prejudice, without taxonomic omissions.”

Not exactly. There are always omissions. That’s why I keep adding taxa to the LRT.  What the LRT represents today is a pretty wide gamut of vertebrate taxa with few to no major clades omitted.

“Therefore, if I were to assemble a matrix of reptiles from scratch, including characters at an order of magnitude similar to his matrix, I would have to obtain a phylogeny similar to his.”

That is a guess. Now turn that ‘if” into a deed.

“The fact that no phylogeny produced in scientific literature produces a phylogeny remotely similar to its own (where the synapsids are internal to the diapsids, and where the pterosaurs result with the squamates, just to mention some of its unusual characteristics) may mean, alternatively that : – All other paleontologists are incompetent and unable to do phylogenetics or that – Peters is incompetent and unable to do phylogenetics.”

Not exactly. Due to the wide gamut of taxa employed by the LRT, wider than in any prior published study, whenever taxa that have never been tested together are now tested together certain hypothetical interrelationships may develop that have been overlooked due to taxon exclusion. Cau’s diphyletic conclusion omits the third possibility: that adding taxa will result in matching tree topologies.

“Who is right? We accept the challenge! Indeed, I accepted it a few years ago, and tried to construct a phylogeny of Reptilia that was assembled from scratch, based exclusively on the characteristics deducible in the literature from the anatomical descriptions of fossil taxa.

As I assembled this matrix, I realized that it was necessary to add characters that were absent in the first few iterations, because I noticed characters that were shared by the new taxa I added and by some of the taxa already present, characters that were not yet in the matrix.

In the end, I found that as I added taxa, the phylogeny tended to produce a small number of scenarios, but nevertheless these were the most frequent with each update of the matrix.

Characters in the LRT are generic, good for almost every situation. I did have to add “no skull” to the headless chordates when they were entered. But otherwise, the original 235 multi-state scores were good to go! And, yes, I am as surprised as you are with regard to their longevity and universality.

In the current version, the assembled matrix has 100 taxa and 355 characters. Although Peters’ rhetoric will complain that 100 taxa are too few, the sample nevertheless took one representative of each major Reptilia group, so it is already an exhaustive sampling of the reptilian anatomical disparity. The matrix also includes soft-part characters (such as integument or internal organs) as it also includes 4 living species representing the 4 main groups of reptiles: chelones, lepidosaurs, crocodiles and birds. It samples species from the Carboniferous to the present, focusing mainly on the Triassic, the period in which pterosaurs appear. The analysis is at the scale of the species (apart from a handful of taxa, however codified on the basis of a few very similar species). I analyzed the matrix both using parsimony and using Bayesian inference which combines morphology and stratigraphy.”

Here Cau is showing his own results from his own taxon list (see below). 100 taxa are likely cherry-picked based on his access to data, even though he doesn’t think so. Did Cau include amphibian-like reptiles? No. Did Cau include fenestrasaurs? Yes, but he did not examine them first-hand. Did Cau examine his results for faults? Clearly no.

“These are the results: 
Pterosauria is monophyletic, and is internal to:
crown-Reptilia (the clade comprising turtles, lepidosaurs and archosaurs)
Sauria (the clade comprising lepidosaurs and archosaurs)
Archosauromorpha (the clade including birds and crocodiles and all taxa closest to them than lepidosaurs)
Archosauria (the clade formed by birds and crocodiles)
pan-Aves (the clade including birds and all taxa closer to them than crocodiles).”

This is the traditional nesting for pterosaurs, but it depends on the omission or poor-scoring of pertinent taxa. Earlier I tested Nesbitt 2011 and found dozens of mistakes in the scores. When corrected Nesbitt’s tree more closely resembles the topology of the LRT.

Added a day later: As noted above, Cau wrote to tell me this morning he would not provide his data file, so I cannot report on scores Cau made to pertinent taxa that I have studied firsthand (Peters 2000) and he has not reported doing so.

“That is, my analysis, based on the anatomical data of 100 reptiles, supports Ornithodira, which is precisely the hypothesis supported by the vast majority of studies on fossil reptiles, and which therefore fully and unquestionably refutes the phylogeny proposed by Peters.”

Readers should know that this analysis comes as news to me. You might think that scientists would share results in cases like this, but they don’t.

“Therefore, returning to the original question of who is incompetent in carrying out phylogenetic analyzes, there are two options: – or I am incompetent in carrying out phylogenetic analyzes and by pure chance I have obtained the same evolutionary scenario resulting from dozens of other paleontologists (who, obviously, are just as incompetent as me) or – Dave Peters is incompetent in doing phylogenetic analysis. You decide which is the most logical explanation.”

Interesting editorial. It does not follow the scientific method, nor was Cau polite enough to send me a .nex file of his results before he launched into this tirade. That shows lack of respect, perhaps contempt, certainly no attempt at reconciliation or consensus. Did Cau realize that Peters 2000, which nested pterosaurs with tanystropheids, Macrocnemus, Langobardisaurus, Cosesaurus, Sharovipteryx and Longisquama came to that conclusion after simply adding these taxa to four previously published analyses? That would have been the first step.

“DISCLAIMER: Contrary to Peters, this post does not intend to revolutionize the phylogeny of reptiles or to propose a new standard hypothesis for pterosaurs (also because, as you have seen, the result coincides with the scenario shared by the majority of paleontologists). If you want to learn something about the evolution of reptiles you have to read the scientific literature published according to the criteria of peer review among serious paleontologists, and not follow the gurus of the net.”

Reading between the lines, Cau has just reported that he has not reviewed Peters 2000, which is in “the scientific literature, published according to the criteria of peer review among serious paleontologists.” The same goes for several other published papers, all of which were peer-reviewed. 

“Rather, here I simply wanted to test Peters’ rhetorical statements playing exactly in his field, to demonstrate that there is no “heretical science” as opposed to a “mainstream science”, but only “good science” which follows a repeatable method, as opposed to a “pseudo-science” that pretends to be scientific and instead is basically a dirty play with data and with the good faith of the readers.”

Cau just told us that he did not follow “a repeatable method”, but instead ran his own experiment with his own list of taxa and not following the literature. Remember when Hone and Benton 2007 said they were going to test Peters 2000 vs. Bennett 1996 in their supertree analysis of pterosaur origins, then in 2009 dropped all reference to Peters 2000 and gave credit to Bennet for both opposing views?

Whenever scientists like Cau claim to be testing using the scientific method in order to discredit someone, check to make sure they really are doing so transparently.

“PS: you may be wondering what is the point of wasting time and energy to deny a pseudo-scientific site that spreads similar ‘cialtronato’ online. Others have talked about it extensively before me, but I think that a scientist has a duty not only to spread scientific knowledge but also to dismantle and demolish mystification and pseudoscience.”

In doing so, make sure you follow every step of the scientific process. As revealed here, Cau did not repeat the experiment, was apparently unaware of the literature and did not communicate is displeasure personally prior to publishing his opinion. His lack of peer review, whether he considers me a peer or not, left him exposed to same criticism he proposed in 2019.

“The site “The Pterosaur Heresies” very subtly spreads a form of pseuso-palentology that harms non-professional readers who may come across that site and think it is a reliable source of paleontological information.”

Don’t trust any source for paleo information. Run the experiment yourself. Find out for yourself. That is good science. Don’t trust anyone who asks you to trust them, as every parent tells their children.

“This is not the case: as I have shown in this post, even trying to “do as he says”, and therefore producing a “ex novo” phylogeny, released from the “official paleontologists” and built independently, without prejudice of any kind in this regard. the evolution of reptiles, drawing directly from the anatomical descriptions of the reptiles, produces a result that not only denies Peters’ absurd phylogenies, but also confirms the robustness and validity of the scenario constructed by the international paleontological community.”

He says “without prejudice”, but did he employ the same taxon list? That’s all I ask. Now let’s look at the cladograms on his blogsite, illegible at their published size. Perhaps clicking on them will expand them to reveal the taxon list.

After clicking on the cladogram: I am disappointed. The legibility does not improve. Seymouriamorpha is the outgroup taxon. That’s not a good start. Eothyris, Paleothyris and Petrolacosaurus and Mesosaurus are the next few taxa. Those cherry-picked taxa are not basal amniotes/reptiles in the LRT, nor are they related to one another. Taxon exclusion by Cau hides the basal dichotomy that splits reptiles at their origin.

Figure 1. Scleromochlus, a basal crocodylomorph

Looking at taxa pertinent to pterosaurs, Cau nests Scleromochlus (Fig. 1) at the base of the Pterosauria. Evidently he does not care that Scleromochlus has vestigial hands and lacks a fifth toe. Cau is not using logic here.

Cau employs Cosesaurus and Sharovipteryx, but did he see these taxa firsthand? I did. Did he create reconstructions so he and we know he fully understands what he is seeing? I did. Did he show the prepubis and pteroid in Cosesaurus? He did not, but you can see them here.

Figure 2. Click to enlarge. The origin of the pterosaur wing and the migration of the pteroid and preaxial carpal. A. Sphenodon. B. Huehuecuetzpalli. C. Cosesaurus. D. Sharovipteryx. E. Longisquama. F-H. The Milan specimen MPUM 6009, a basal pterosaur.

Some other odd cousins arise in Cau’s cladogram:

1. Parasuchus, the phytosaur, nests with Alligator, the crocodylomorph.
2. The ‘rib’ glider Icarosaurus is in the turtle lineage.
3. Long-snouted aquatic Mesosaurus nests with short-snouted terrestrial Acleistorhinus.

Like Darren Naish,
Andrea Cau worked hard for his PhD in the hope of making discoveries in paleontology. When writing papers, the review process takes months that might result in a rejection.

By contrast, 
the LRT can add taxa and make comment within 24 hours, making discoveries without the cost of a PhD. Several times the LRT recovered hypothetical relationships years before they were published by PhDs. You can see that list here. Cau dismissed all of those discoveries as lucky. So…

1. Cau is not without prejudice.
2. Cau did not check the results of his cladogram for logic
3. Cau did not show his tracings or make reconstructions. Scores are suspect.
4. Cau did not employ taxa pertinent to the basal reptile dichotomy
5. Cau did not send me his manuscript for peer review, so this is the review he never asked for.